Generating and filtering navigational maps
Described are techniques for generating, updating, and using sensor-based navigational maps. An input map is generated based on sensor data captured by a first sensor of a first vehicle. The input map is filtered based on one or more criteria to generate a filtered map corresponding to a three-dimensional representation of a route traveled by the first vehicle. The filtering can be based on detecting features using sensor data captured by a second sensor of the first vehicle. The one or more criteria can include object classes, distance criteria, and/or other criteria relating to attributes of features in the sensor data captured by the first sensor and/or the sensor data captured by the second sensor. The filtered map can be stored for transmission to a second vehicle, for use in determining a location of the second vehicle while the second vehicle is traveling along the same route.
1. A method for generating a map, the method comprising:
generating, by a computer system and based on sensor data captured by a first sensor of a first vehicle, an input map corresponding to a three-dimensional representation of a route traveled by the first vehicle;
detecting, by the computer system, objects represented in the sensor data, wherein the detecting comprises assigning a corresponding class label to features associated with the same object, each class label being selected from a set of predefined object classes;
generating, by the computer system, a bounding box around each detected object;
filtering, by the computer system and based on one or more criteria, the input map to generate a filtered map, wherein:
the filtering comprises automatically removing features that satisfy the one or more criteria,
the one or more criteria identify objects that are irrelevant to determining a location of a second vehicle with reference to the filtered map, and
the one or more criteria comprise at least one object class such that at least some features are removed based on having been assigned a class label corresponding to the at least one object class and further based on belonging to the same bounding box; and
transmitting, by the computer system, the filtered map to the second vehicle, wherein the filtered map is processed by a navigation system of the second vehicle to determine a location of the second vehicle while the second vehicle is traveling along the same route.
2. The method of claim 1 , wherein the at least one object class comprises at least one of vehicles, pedestrians, animals, or plants.
3. The method of claim 1 , wherein the one or more criteria comprise at least one attribute of an object to be removed, the at least one attribute including at least one of age, length, width, height, speed, or shape.
4. The method of claim 1 , wherein the one or more criteria comprise a distance from the first vehicle, the distance being a threshold distance beyond which features are to be removed.
5. The method of claim 1 , wherein the filtered map indicates edges of a multi-lane road along the route, and wherein the filtered map excludes lanes within the multi-lane road.
6. The method of claim 1 , wherein the first sensor is a LIDAR sensor, wherein the input map comprises a point cloud captured by the LIDAR sensor, and wherein the filtering comprises removing points that satisfy the one or more criteria.
7. The method of claim 6 , wherein detecting objects represented in the sensor data further comprises:
identifying features belonging to the same object using the point cloud together with additional sensor data captured by a second sensor of the first vehicle, wherein the additional sensor data is captured contemporaneously with the point cloud, and wherein the second sensor is a camera, a radar sensor, or another LIDAR sensor.
8. The method of claim 7 , wherein the second sensor is a camera, and wherein generating a bounding box around each detected object comprises:
determining two-dimensional boundaries of objects in an image captured by the camera; and
mapping the two-dimensional boundaries to corresponding regions of the point cloud to form three-dimensional bounding boxes around points in the point cloud.
9. The method of claim 1 , wherein the filtering comprises:
performing a first filtering operation in which features corresponding to ground reflections and moving objects are removed from the input map, the moving objects being removed through point motion classification; and
performing a second filtering operation after the first filtering operation, the second filtering operation removing features that have been assigned a class label corresponding to the at least one object class.
10. The method of claim 1 , wherein the filtered map is transmitted to the second vehicle through a wireless connection while the second vehicle is traveling along the same route, and wherein to determine the location of the second vehicle, features included in the filtered map are matched to corresponding features in sensor data captured by a sensor of the second vehicle.
11. A system comprising:
one or more processors; and
a memory storing instructions that, when executed by the one or more processors, cause the one or more processors to:
generate, based on sensor data captured by a first sensor of a first vehicle, an input map corresponding to a three-dimensional representation of a route traveled by the first vehicle;
detect objects represented in the sensor data, wherein the one or more processors are configured to detect objects through assigning a corresponding class label to features associated with the same object, each class label being selected from a set of predefined object classes;
generate a bounding box around each detected object;
filter, based on one or more criteria, the input map to generate a filtered map, wherein:
to filter the input map the one or more processors are configured to automatically remove features that satisfy the one or more criteria,
the one or more criteria identify objects that are irrelevant to determining a location of a second vehicle with reference to the filtered map, and
the one or more criteria comprise at least one object class such that at least some features are removed based on having been assigned a class label corresponding to the at least one object class and further based on belonging to the same bounding box; and
transmit the filtered map to the second vehicle, wherein the filtered map is usable by a navigation system of the second vehicle to determine a location of the second vehicle while the second vehicle is traveling along the same route.
12. The system of claim 11 , wherein the at least one object class comprises at least one of vehicles, pedestrians, animals, or plants.
13. The system of claim 11 , wherein the one or more criteria comprise at least one attribute of an object to be removed, the at least one attribute including at least one of age, length, width, height, speed, or shape.
14. The system of claim 11 , wherein the one or more criteria comprise a distance from the first vehicle, the distance being a threshold distance beyond which features are to be removed.
15. The system of claim 11 , wherein the filtered map indicates edges of a multi-lane road along the route, and wherein the filtered map excludes lanes within the multi-lane road.
16. The system of claim 11 , wherein:
the first sensor is a LIDAR sensor;
the input map comprises a point cloud captured by the LIDAR sensor;
the one or more processors are configured to identify features belonging to the same object using the point cloud together with additional sensor data captured by a second sensor of the first vehicle;
the additional sensor data is captured contemporaneously with the point cloud captured by the LIDAR sensor; and
the second sensor is a camera, a radar sensor, or another LIDAR sensor.
17. The system of claim 16 , wherein the second sensor is a camera, and wherein to generate a bounding box around each detected object, the one or more processors are configured to:
determine two-dimensional boundaries of objects in an image captured by the camera; and
map the two-dimensional boundaries to corresponding regions of the point cloud to form three-dimensional bounding boxes around points in the point cloud.
18. The system of claim 11 , wherein to filter the input map, the one or more processors are configured to:
perform a first filtering operation in which features corresponding to ground reflections and moving objects are removed from the input map, the moving objects being removed through point motion classification; and
perform a second filtering operation after the first filtering operation, the second filtering operation removing features that have been assigned a class label corresponding to the at least one object class.
19. The system of claim 11 , wherein the filtered map is transmitted to the second vehicle through a wireless connection while the second vehicle is traveling along the same route, and wherein the navigation system of the second vehicle is configured to determine the location of the second vehicle through matching features included in the filtered map to corresponding features in sensor data captured by a sensor of the second vehicle.
20. A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors of a computer system, cause the one or more processors to perform the following:
generating, based on sensor data captured by a first sensor of a first vehicle, an input map corresponding to a three-dimensional representation of a route traveled by the first vehicle;
detecting, by the computer system, objects represented in the sensor data, wherein the detecting comprises assigning a corresponding class label to features associated with the same object, each class label being selected from a set of predefined object classes;
generating, by the computer system, a bounding box around each detected object;
filtering, based on one or more criteria, the input map to generate a filtered map, wherein:
the filtering comprises automatically removing features that satisfy the one or more criteria,
the one or more criteria identify objects that are irrelevant to determining a location of a second vehicle with reference to the filtered map, and
the one or more criteria comprise at least one object class such that at least some features are removed based on having been assigned a class label corresponding to the at least one object class and further based on belonging to the same bounding box; and
transmitting the filtered map to the second vehicle, wherein the filtered map is usable by a navigation system of the second vehicle to determine a location of the second vehicle while the second vehicle is traveling along the same route.